LPC based low dimensional features for object classification

Abstract

Object classification in both images and videos is an important task within the field of computer vision. The process of classifying objects into predefined and semantically meaningful categories using its features is called object classification. Many researchers are working in this area to improve the accuracy of classification and to reduce the dimension of features extracted which are used for classifying the objects. In this paper, we propose Linear Predictive Coding(LPC) based signal approximation on the Tensor features which reduces the dimension of the feature by removing the redundancies in the feature set, so that the accuracy is increased with less computation time. Deep Neural Network (DNN) which is a type of artificial intelligence that could solve complex perceptual problems as fast as human brain is used in this work to classify the objects in videos and images. In the proposed method we employ the combination of Scale Invariant Feature Transform (SIFT) and Tensor features of reduced dimensions. Simulation results illustrate that the proposed model produces more accurate results than many existing methods.